Writing prompts that work
1. Writing prompts that work
Asking AI to "analyze this data" rarely produces something useful. That's not a model problem — it's a prompt problem. The output an AI gives you is influenced by the input you give it, and getting prompts right is the difference between real analysis and generic noise.2. The evidence
Research from Anthropic — the team behind Claude — found that the sophistication of Claude's responses tends to match the sophistication of the user's input. Put effort into your prompt and the AI rises to meet you. Skip it, and you get something generic.3. GCSE framework
That's where the GCSE framework comes in. It gives you a clear structure and contents for the prompts you write.4. G is for Goal
G stands for Goal. What do you want the AI to produce? What is the task? "Analyze my data" isn't a goal — it's a wish. "Tell me which coffee drink is the most popular across all our stores" — that's a goal. Be specific about the output you want.5. C is for Context
C stands for Context — the why. The AI knows nothing about you, your business, or how your data is used. And even with familiar data — sales numbers, social media metrics — the specifics matter. Different companies define "active user" differently: people who log in every day, every month, or every six months. The AI cannot guess that. You have to tell it.6. S is for Scope
S stands for Scope — the boundaries. What should or shouldn't be in the answer? A specific time period? Certain segments to exclude? Or a particular methodology? Scope keeps the AI from wandering off into territory you don't need.7. E is for Examples
E stands for Examples. This is often the most powerful piece of the framework. Show the AI what good output looks like — the format, the structure, the level of detail — and it'll match that pattern. One concrete example beats three paragraphs of description every time.8. GCSE in action
Let's see this in action. We're analysts for The Daily Grind, a chain of coffee stores. A loose prompt like "Analyze our data" gets us an equally loose response. Now let's add each piece. Our goal: find the most popular coffee drink across all stores. Our context: we're planning new products and want decisions informed by what's popular and profitable. Our scope: in-store transactions only, excluding mobile orders. Our example: classify drinks on two axes — high, medium, low sales against high, medium, low profits. That gives you the full prompt — assembled and ready to send. The AI now knows what you want, why you want it, what to leave out, and what the output should look like. Same data, completely different result.9. Let's practice!
Time to put GCSE to work in your own prompts!Create Your Free Account
or
By continuing, you accept our Terms of Use, our Privacy Policy and that your data is stored in the USA.